Windows VPS: performance optimization

Quick audit: Task Manager + PerfMon

1. Open Task Manager → Performance and view the load on CPU, Memory, Disk, Ethernet. For a quick disk check, focus on “Active time” and “Average response time.”

2. Run perfmon.msc → Performance Monitor → + and add the basic counters:

• CPU: % Processor Time, % Privileged Time, % User Time
• Memory: Available MBytes, Pages/sec
• Disk: Avg. Disk sec/Transfer, Disk Transfers/sec, Avg. Disk Queue Length
• Save in Data Collector Set for 1–5 minutes to get a report.

Clean up and tidy up

Disk Cleanup (cleanmgr): Explorer → C: drive → Properties → Disk Cleanup → Clean up system files. Select Windows Update Cleanup, temporary files, recycle bin, and WER.

Autostart:

In Windows Server 2019/2022, enable Storage Sense to automatically clean up Temp/Downloads/Recycle Bin on a schedule.

Disks: objective speed measurement

Install DiskSpd and run the test (example — 40% write, 60% read; 64K block; 8 threads; 32 queue; 180 sec):

See IOPS, AvgLat (ms), and MiB/s; stable disk latency in prod — <10–25 ms if possible.

If disk graphs are not visible in Task Manager on older systems: run diskperf -y and restart Task Manager.

Network: removing oversights (RSC/VMQ/RSS/LSO)

On a number of configurations, disabling some offload functions and carefully configuring TCP helps:

After each step, measure using PerfMon/iperf/SMB copying to avoid “fixing one thing and breaking another.”

Antivirus: exceptions for “hot” paths

To prevent the scanner from throttling I/O (SQL, log folders, caches), add exceptions in Microsoft Defender:

Parameters and commands — in the official Defender documentation.

Memory and pagefile

Keep 15–20% of RAM free for the file cache. Pagefile — “System-selected” or fixed (min=RAM, max≈1.5×RAM, if full crash dumps are needed).

SMART/drive health

Check for SSD/HDD predictive errors:

If you see media degradation, plan your migration before an incident occurs.

Checklist

• PerfMon reports before/after, alerts on CPU/disk/network.
• Disk Cleanup/Storage Sense configured.
• DiskSpd shows acceptable latency.
• Network offloads checked (RSC/VMQ/RSS/LSO).
• Defender exceptions for “hot” paths/processes.
• SMART monitoring without warnings.

Table of Contents